Thousands of unnamed workers in India provide the critical data annotation required to train every major AI model. This invisible labor force lacks formal recognition and systemic support. Their ability to label complex datasets directly dictates the accuracy of LLMs. Practitioners must now account for this human bottleneck in the AI supply chain.